Deceased Humpback Whale - Joshua Meza-Fidalgo, 2020

Deceased Humpback Whale - Joshua Meza-Fidalgo, 2020

Rationale and Research Questions

Marine mammal and sea turtle strandings can be unusual events or indicators of problems in our ocean ecosystems. Seeing patterns in the strandings of marine mammals and sea turtles can be indicative of more complex issues that may not be transparent. Some examples of these issues could be viral outbreaks in lower trophic level populations or toxic algal blooms that kill off all sea life in the area. Because humans demand and consume seafood, understanding the problems that arise are essential to preventing the consumption of contaminated seafood along with the management of fisheries. By studying marine mammal and sea turtle strandings, we get insight on the health of our oceans.

Our objective is to analyze animal strandings over the years to see if there any trends in the quantity of strandings for each family (cetaceans, odontocetes, mysticetes, and pinnipeds). We have decided to look at the total number of strandings per year for all families, the changes in strandings over the years for each family and the seasonality differences across all years for each family.

Dataset Information


Table 1: Dataset Information
Detail Description
Data Source OBIS-SEAMAP - Mystic Aquarium
Retrieved From https://seamap.env.duke.edu/dataset/945
Variables Used Family, Common Name, Species Name, Date, Year, Month, Latitude, Longitude, Total Strandings, Strandings by Species
Data Range March 18, 1976 to December, 29, 2011


The dataset that we chose to work with is the Mystic Aquarium’s marine mammal and sea turtle stranding data that started in 1976 until 2011. However, there was only one stranding record in 1976, so that point was excluded from our analysis. The rest of the data started in 1990 and was concluded in 2011. It contains the number of whale (odontocetes and mysticetes), pinniped (seals and sea lions), and sea turtle strandings that occurred along the coasts of Connecticut, Rhode Island and Fishers Island, New York, USA. There were a total of 1140 strandings; within these strandings, 155 were odontocetes, 50 were mysticetes, 679 were pinnipeds, and 256 were sea turtles.

The dataset has a latitude that ranges from 41.00 to 41.86 while the longitude ranges from -73.65 to -71.10. The data was provided using the datum WGS 1984 (4326) and we performed a data transformation into UTM 19 (32619) for geospatial analysis. The data was wrangled to contain only the variables that we were interested in using. We then selected parts of the data based off of the Common Name of the species to create datasets that were easier for us to use in our analysis. This helped us divide the data into four groups: odontocetes, mysticetes, pinnipeds, and sea turtles. We initially looked at total yearly strandings for trends in the data for each family. After we analyzed the data on a monthly basis to help give insight to seasonal stranding trends for each family.


Exploratory Analysis

We initially decided that it would be best to look at the data holistically to get a better feel for the trends in the data. We first analyzed the total number of strandings per each family, as seen in Figure 1. To better see the data, we divided and summed the strandings of each family for each year from 1990 to 2011, as seen in Figure 2. To see if there could be seasonal trends for the entire data set, the data was grouped by month for each family from the years 1990 to 2011, as seen in Figure 3. Finally, to have a geospatial understanding of our data, Map 1 was created.

Figure 1: Total Strandings per Family


Figure 2: Total Strandings per Year


Figure 3: Total Strandings per Month


Map 1: Geospatial Exploration of each Family


Analysis

Because we decided to divide the data into groups by their families, we conducted separate analysis for each of the families. Our research questions are the following:

1. Are there years or months where the number of strandings for each family are significantly different?
2. Are there trends in each of the family strandings based off this dataset?

We utilized GLMs, MannKendall, and Seasonal MannKendall in our analysis. Due to our data being count data rather than continuous, we used a poisson regression GLM.

1. Pinnipeds:

Pinniped Data Exploration

Figure 4: Total Pinniped Strandings per Year by Species


Figure 5: Total Pinniped Strandings per Month by Species


Figure 6: Pinniped Strandings per Year


Pinniped data exploration indicates a possible overall increasing trend in pinniped strandings and a possible monthly/seasonal trend. There is some fluctuation in number of total yearly strandings both for the family as a whole and at the species level, although a more detailed species level analysis is beyond the scope of this project. Harp and harbor seals make up most of the strandings each year. Most yearly strandings occur in the late winter and spring time, particularly in March, with very few in the late summer through the early winter. Springtime coincides with the pupping seasons for many seal species and could be a factor in this. Pinniped strandings are most numerous throughout the coastline of Rhode Island and the eastern coastline of Connecticut, as seen in Map 2 below.

Map 2: Geospatial Exploration of Pinniped Strandings


Pinniped Stranding Statistical Analysis

The following are the null and alternative hypothesis for the annual and monthly statistical analysis using Poisson Regression (GLM):

Null Hypothesis 1: There is no effect of year on the total number of pinniped strandings.
Alternative Hypothesis 1: There is an effect of year on the total number of pinniped strandings.

Null Hypothesis 2: There is no effect of month on the total number of pinniped strandings.
Alternative Hypothesis 2: There is an effect of month on the total number of pinniped strandings.

Yearly Analysis:
We reject the null hypothesis that there is no effect of year on the total number of pinniped strandings. 18 of the 21 documented years were statistically significant with p-values ranging from 9.07e-09(2001) to 0.03(1999). Only the first three years of data, 1991-1993 had p-values above 0.05. The null deviance of yearly model was 1.36e+02 on 21 degrees of freedom, and residual deviance of 1.31e-14

Monthly Analysis:
We reject the null hypothesis that there is no effect of month on the total number of pinniped strandings. 11 of the 12 months had p-values of less than 0.05, with July being the exception. The p-values of the statistically significant month ranged from 1.77e-15(March) to 0.02(June). The null deviance for the monthly model was 5.95e+02 on 11 degrees of freedom and a residual deviance of -3.55e-15.

This information can lead to further study as to why years and months differ and develop ways to predict and/or mitigate particularly bad years and/or months for pinniped strandings.

Pinniped Geospatial Analysis


Map 3: Pinniped Distance (m) from Mean Stranding Point


By finding the “mean” stranding location based on the all of the data points it can be determined how far each stranding point is from the mean. By determining that 425 out of 679, or 62.5 percent of, pinniped strandings occur within 40km of the mean stranding location, what makes this area a hotspot can be studied and hopefully also determined. This data also could allow for reallocation of stranding resources and personnel to this range in order to put them to the most efficient use possible. Since only 54, or 8 percent of, pinniped strandings occur further than 60km from the stranding mean, resources for study and rescue can be allocated away from those areas. A table containing the distance of the all of the data points from the “mean” stranding location can be found in Table 2 below.


Table 2: Number of Pinniped Strandings per Distance (m) from the Mean
Distance (m) Number of Strandings
20,000 - 40,000 425
40,000 - 60,000 200
60,000 - 80,000 15
80,000 - 100,000 11
100,000 - 120,000 7
120,000 - 140,000 11
140,000 - 160,000 7
greater than 160,000 3


Pinniped Temporal Analysis

The following are the null and alternative hypothesis for the yearly and monthly time series:

Null Hypothesis: The yearly and/or monthly/seasonal pinniped stranding data is stationary.

Alternative Hypothesis: The yearly and/or monthly/seasonal pinniped stranding data is not stationary.

Figure 7: Pinniped Monthly Decomposition


A Mann Kendall test was run for the yearly time series. Based on the results, we would reject the null hypothesis that yearly pinniped strandings are stationary with a p-value of 1.77e-14.

A seasonal Mann Kendall test was run for the monthly time series. Based on the results, we reject the null hypothesis that monthly pinniped strandings are stationary, with a p-value of 7.80e-6. The decomposition indicates a general upward trend as well as seasonality, which fits with the patterns that were seen in the exploratory analysis (Figure 7).


2. Odontocetes:

Odontocete Data Exploration

Figure 8: Total Odontocete Strandings per Year by Species


Figure 9: Total Odontocete Strandings per Month by Species


There appears to be an increasing trend in the total number of odontocete strandings as the years progress (Figure 8). The odontocete strandings have the lowest amount in 1993 and the highest in 2011, where the count gradually increases over the years. The different species that are stranded throughout the years do not appear to have a noticeable trend. When looking at the strandings per month in Figure 9, the months with the most strandings are during March, May, and July. There may be possible seasonal trends, where the spring and early summer see a greater number of strandings than the other months due to calving season. When looking at Figure 10 below, you can see the trend in the number of strandings of odontocetes over the years, and how they are increasing. All of these strandings can be seen in the Map 4 below.


Figure 10: Odontocete Strandings per Year


Map 4: Geospatial Exploration of Odontocete Strandings


Odontocete Stranding Statistical Analysis

The following are the null and alternative hypothesis for the annual and monthly statistical analysis using Poisson Models (GLM):

Null Hypothesis 1: There is no effect of individual year on the total number of odontocete strandings.
Alternative Hypothesis 1: There is an effect of year on the total number of odontocete strandings.

Null Hypothesis 2: There is no effect of month on the total number of odontocete strandings.
Alternative Hypothesis 2: There is an effect of month on the total number of odontocete strandings.

Yearly Analysis:
We reject the null hypothesis that there is no effect of year on the total number of odontocete strandings. Of all the years in the odontocete strandings, only 2 years are statistically significant; 1990 and 2011 were both statistically significant with p-values of 1.14e-05 and 0.0198 respectively. The null deviance was had a value of 4.0129e+01 and the residual deviance had a value of. -2.1867e-21.

Monthly Analysis:
We reject the null hypothesis that there is no effect of month on the total number of odontocete strandings. There were four months in which the odonotocete strandings were statistically significant: January, March, May, and June. The most statistucally significant months were January and May with p-values of 4.35e-11 and 0.00858 respectively. March had a p-value of 0.02389 and June had a p-value of 0.03344. The null deviance was 3.7282e+01 and the residual deviance was -3.5527e-15.

This indicates that there is a difference in the number of odontocete strandings across the years and months of the study. This could be explained by increase boat and shipping traffic as the years progress that contribute to increased ocean noise and interference with their communication that have led them to become stranded. Seasonality may not have an effect because of the consistency of boat and ship traffic throughout each year.


Odontocete Geospatial Analysis


Map 5: Odontocete Distance (m) from Mean Stranding Point


Knowing where the “mean” stranding location is could allow for a study of that area in order to determine why odontocete strandings are so prominent on this area and allow for the allocation of more recovery resources and personnel to that area to compensate for the higher levels of strandings. Much of the strandings occurred from Charlestown to Newport, in the Rhode Island Sound.


Table 3: Number of Odontocete Strandings per Distance (m) from the Mean
Distance (m) Number of Odontocete Strandings
less than 40,000 125
40,000 - 60,000 20
60,000 - 80,000 1
80,000 - 100,000 0
100,000 - 120,000 3
120,000 - 140,000 2
140,000 - 160,000 1
160,000 - 180,000 2
grater than 180,000 1


Of the 155 odontocete strandings used in our analysis, 125 of them were within 40,000m of mean latitude and longitude stranding location. This was followed by 20 strandings that were within 40,000m to 60,000m from the mean stranding location. The remaining 10 strandings were scattered farther than 60,000m, and there was one stranding that was greater than 180,000m from the mean stranding location. These values can be seen in Table 3 above. Most of the strandings were centered near Newport, in the Rhode Island Sound. There were also many strandings that were located up in Narragansett Bay. The increased stranding areas in Rhode Island Sound may be due to anthropogenic effects such as higher boat traffic and increased amounts of fishing. It may also be due to the area having an abundance of food for the odontocetes, which push them closer to shore. This area is densely populated, so the stranding count may be higher because of closer monitoring. Further research will be needed to determine what may contribute to odontocetes stranding from Charlestown to Newport, in the Rhode Island Sound.


Odontocete Temporal Analysis

The following are the null and alternative hypothesis for the yearly and monthly time series:

Null Hypothesis: The yearly and/or monthly/seasonal odontocete stranding data is stationary.

Alternative Hypothesis: The yearly and/or monthly/seasonal odontocete stranding data is not stationary.

Figure 11: Odontocete Monthly Decomposition


A Mann Kendall test was run for our yearly time series. Based on the results, we reject the null hypothesis and say that odontocete yearly stranding data is not stationary due to a p-value of 4.4296e-13.

A seasonal Mann Kendall test was run for our monthly time series. Based on the results, we reject the null hypothesis and say that odontocetes monthly stranding data is not stationary due to a p-value of 1.3611e-07. This can also be seen in our decomposition where there appears to be a seasonal trend in the data over the years (Figure 11).


3. Mysticetes

Mysticete Data Exploration

Figure 12: Total Mysticete Strandings per Year by Species


Figure 13: Total Mysticete Strandings per Month by Species


There appears to be an increasing trend in the number of mysticete strandings in the later years of the study. The year with the highest number of strandings was 2004 and the second highest was 2009. There was also lots of variation with the species of that were stranded in each year. When looking at Figure 13, there also appears to be a seasonal trend based on the number of strandings per each month, where the months of June and July have a much greater amount than the rest of the year. There may be possible seasonal trends, where the early summer see a greater number of strandings than the other months due to migration. The area may be a common feeding ground during their seasonal migrations along the east coast.


Figure 14: Mysticete Strandings per Year


When looking at Figure 14 above, you can see the trend in the number of strandings of mysticetes over the years is increasing, but not by much. It appears that the number of strandings is consistent until 2000, where the numbers increase before dropping down again in 2009. All of these strandings can be seen in Map 6 below.


Map 6: Geospatial Exploration of Mysticete Strandings


Mysticete Stranding Statistical Analysis

The following are the null and alternative hypothesis for the annual and monthly statistical analysis using Poisson Models (GLM):

Null Hypothesis 1: There is no effect of year on the total number of mysticete strandings.
Alternative Hypothesis 1: There is an effect of year on the total number of mysticete strandings.

Null Hypothesis 2: There is no effect of month on the total number of mysticete strandings.
Alternative Hypothesis 2: There is an effect of month on the total number of mysticete strandings.

Yearly Analysis:
We reject the null hypothesis that there is no effect of year on the total number of mysticete strandings due to none of the p-values in each year being below 0.05. This means that overall, there is no change in the number of strandings over the years of the study. The null deviance had a p-value of 2.1360e+01, and the residual deviance had a p-value of 4.1223e-10.

Monthly Analysis:
We fail to reject the null hypothesis that there is no effect of month on the total number of mysticete strandings. There were two months in which the mysticete strandings were statistically significant: May and June. The most statistically significant month was May with a p-value of 0.00874 while June had a p-value of 0.01345. The null deviance had a p-value of 5.7723e+01 the residual deviance had a p-value of 3.0330e-10.

This indicates that there is no difference in the number of mysticete strandings across the years, but there is a difference in the months of the May and June study. Seasonality appears to affect the number of mysticete strandings as there are more occurances druring May and June than any other month while there is not a significant difference in the number of strandings in different years. The seasonal increase in strandings may be due to seasonal migration to feeding grounds, but further research would be needed to determine the cause.


Mysticete Geospatial Analysis


Map 7: Mysticete Distance (m) from Mean Stranding Point


Knowing where the “mean” stranding location is could allow for a study of that area in order to determine why mysticete strandings are so prominent on this area and allow for the allocation of more recovery resources and personnel to that area to compensate for the higher levels of strandings. Much of the strandings occurred from Stonington to Newport, in the Rhode Island Sound with one stranding that was farther away.


Table 4: Number of Mysticete Strandings per Distance (m) from the Mean
Distance (m) Number of Strandings
less than 40,000 44
40,000 - 60,000 5
60,000 - 80,000 0
80,000 - 100,000 0
100,000 - 120,000 0
120,000 - 140,000 0
140,000 - 160,000 1


Of the 50 mysticete strandings used in our analysis, 44 of them were within 40,000m of mean latitude and longitude stranding location. This was followed by 5 strandings that were within 40,000m to 60,000m from the mean stranding location. The remaining stranding was located between 140,000m to 160,000m away from the mean stranding location. These values can be seen in Table 4 above. Most of the strandings were centered near Newport, in the Rhode Island Sound. There were also some strandings that were located up in Narragansett Bay. The increased stranding areas in Rhode Island Sound may be due to anthropogenic effects such as higher boat traffic and increased amounts of fishing. It may also be due to the area having an abundance of food for the mysticetes, which push them closer to shore. This area is densely populated, so the stranding count may be higher because of closer monitoring. Further research will be needed to determine what may contribute to mysticetes stranding from Charlestown to Newport, in the Rhode Island Sound.


Mysticete Temporal Analysis

The following are the null and alternative hypothesis for the yearly and monthly time series:

Null Hypothesis: The yearly and/or monthly/seasonal mysticete stranding data is stationary.

Alternative Hypothesis: The yearly and/or monthly/seasonal mysticete stranding data is not stationary.


Figure 15: Mysticete Monthly Decomposition


A Mann Kendall Test was run for our yearly time series. Based on the results, we reject the null hypothesis and say that mysticete yearly stranding data is not stationary due to a p-value of less than 2.22e-16.

A seasonal Mann Kendall Test was run for our yearly time series. Based on the results, we reject the null hypothesis and say that mysticetes monthly stranding data is not stationary due to a p-value of 0.014425. This can also be seen in our decomposition where there appears to be a seasonal trend in the data over the years (Figure 15).


4. Sea Turtles

Sea Turtle Data Exploration


Figure 16: Total Sea Turtle Strandings per Year by Species


Figure 17: Total Sea Turtle Strandings per Month by Species


Figure 18: Sea Turtle Strandings per Year


Exploratory analysis of the sea turtle stranding data set by year shows higher amounts of strandings in the early 1990s and late 2000s, with the highest amount in 1995 (Figure 16). The most prominent species throughout our study period appear to be leatherback sea turtles, followed by loggerhead sea turtles. Through monthly data exploration, strandings peak from July to September and occur less frequently the remainder of the year (Figure 17).


Map 8: Geospatial Exploration of Sea Turtle Strandings


Sea Turtle Statistical Analysis

The following are the null and alternative hypothesis for the annual and monthly statistical analysis using Poisson Models (GLM):

Null Hypothesis 1: There is no effect of individual year on sea turtle strandings.
Alternative Hypothesis 1: There is an effect of individual year on sea turtle strandings.

Null Hypothesis 2: There is no effect of month on sea turtle strandings.
Alternative Hypothesis 2: There is an effect of month on sea turtle strandings.

Yearly Analysis:
We reject the null hypothesis that individual year has no effect on sea turtle strandings. The sea turtle strandings in 9 of the 21 years are not significantly different (p-values: 0.147 - 0.706). 1995 was the most significantly different (p-value: 0.000236). 1991 (p-value: 0.0056), 1993 (p-value: 0.0078), 2008 (p-value: 0.0022), 2010 (p-value: 0.0016), and 2011 (p-value: 0.003) were the next most statistically different years. The null deviance of this model is 7.4606e+01 and the residual deviance is -4.8850e-15.

Monthly Analysis:
We reject the null hypothesis that month has no effect on sea turtle strandings. Of the 12 months, 5 of months were significantly different. These months include May (p-value: 0.000269), June (p-value: 3.98e-06), July (p-value: 1.91e-05), August (p-value: 0.004106), and September (p-value: 0.049935). This indicates that sea turtle strandings are concentrated to the summer months in our study area. The null deviance of this model is 4.9235e+02 and the residual deviance is 3.0332e-10.


Sea Turtle Geospatial Analysis


Map 9: Sea Turtle Distance (m) from Mean Strandign Point


Of the 256 sea turtle strandings used in our analysis, 113 of them occured within 20,000 - 40,000m of the mean stranding location, followed by 92 strandings within 20,000m of the mean location (Table 5). These points were in the areas around Charlestown, RI, Newport, RI and New Shoreham, RI. Some sea turtle strandings were reported inland, which could be due to tides/currents bringing the animal upstream into rivers (Map 9).This high concentration of strandings can indicate that these areas may be sea turtle stranding hotspots. This could be due to various factors, such as the flow of oceanic currents, anthropogenic impacts (such as higher fishing pressure), or the fact that there may be more manpower employed in these areas that could make stranding detection easier. Further research is needed to determine how these underlying factors may be contributing to sea turtle strandings in these areas.


Table 5: Number of Sea Turtle Strandings per Distance (m) from the Mean
Distance (m) Number of Strandings
less than 20,000 92
20,000 - 40,000 113
40,000 - 60,000 32
60,000 - 80,000 2
80,000 - 100,000 4
100,000 - 120,000 2
120,000 - 140,000 6
140,000 - 160,000 4
greater than 160,000 1


Sea Turtle Temporal Analysis

The following are the null and alternative hypothesis for the yearly and monthly time series:

Null Hypothesis: The yearly and/or monthly/seasonal turtle stranding data is stationary.

Alternative Hypothesis: The yearly and/or monthly/seasonal turtle stranding data is not stationary.


Figure 19: Sea Turtle Decomposition


A Mann Kendall Test was run for our yearly time series. Based on the results, we reject the null hypothesis that yearly turtle strandings are stationary with a p-value =< 2.22e-16.

A seasonal Mann Kendall Test was run for our monthly time series. Based on the results, we fail to reject the null hypothesis that monthly turtle strandings are stationary with a p-value = 0.270.


Summary and Conclusions

1. Pinnipeds:

Both graphing and a time series analysis indicate that there is trend in the yearly totals of pinniped strandings and this trend is increasing. Additionally, 85.7 percent of the years were statistically significant for pinniped strandings. The monthly time series indicated that there was a slight upward monthly trend along with seasonality, which could be attributed to pupping seasons. 11 of the 12 months were statistically significant for strandings. Realizing there are trends is the first step in studying what is causing these trends and to determine if there are ways to stop or manage them.

62.5 percent of pinniped strandings happened within 40km of the “mean stranding point.” By knowing where and when pinnipeds are most likely to strand, the environmental factors of those places and times can be studied in order to determine the conditions that cause strandings. This knowledge could aid in stranding prevention and/or mitigation measures. Further study could also be conducted at the species level to determine what if any differences occur.

2. Odontocetes:

Our analysis indicated there were was change in the total number of strandings over the years of the study. The year 1990 and 2011 were statistically significant, which meant that those years were different than the rest of the study. There was also to be an increase in the number of strandings during the months of January, March, May, and June. This was further shown in our time series where the strandings were not stationary over the years or the months. The increasing stranding over the years may be in indicator that there is an issue that needs to be address through further research. The increased strandings during the earlier months may be contributed to calving season or seasonal migrations. Again, further research would be needed to determine the cause.

We found that 80.6% of our strandings were within 40,000m of the mean stranding point and 93.5% fell within 60,000m of the mean stranding point. This is an indicator that majority of the strandings were located in very similar areas which may show that area is a hazard to odonotocetes. This is another area that needs have further research in attempt to understand why the area is a common stranding point for odontocetes and what can be done to prevent these strandings.

3. Mysticetes:

Our analysis indicated there were was no change in the total number of strandings over the years of the study. None of the years were statistically significant, which meant that overall, there was no major difference between the number of strandings over the study period. There was an increase in the number of strandings during the months of May and June. This was further shown in our time series where the strandings were not stationary over the years or the months. The consistent number of strandings over the years indicate that there may not be any changes in the environment over the study period. The increased strandings during the months of May and June may be contributed to seasonal migrations. Again, further research would be needed to determine the cause.

We found that 88% of our strandings were within 40,000m of the mean stranding point and 98% fell within 60,000m of the mean stranding point. This is an indicator that majority of the strandings were located in very similar areas which may show that area is a hazard to mysticetes. This is another area that needs have further research in attempt to understand why the area is a common stranding point for mysticetes and what can be done to prevent these strandings.

4. Sea Turtles:

80% of sea turtles strandings all occurred within 40km of the mean stranding location. This can indicate that this area could be a hot spot for sea turtle strandings, whether it be due to high abundance of sea turtles in the area or due to oceanic currents. 47.6% of years were statistically significant for sea turtle strandings. 41.7% of months were statistically significant, which are the May - September. This monthly breakdown corresponds with the timing of the sea turtle nesting season and when there is likely to be a higher abundance of them in the area. However, time series analysis indicated that there was a slight significance in year and there was not a significant seasonal component for sea turtle strandings. While our data exploration showed that most of the strandings occurred in the summer months, this could be due to certain months of certain years having disproportionately higher amounts of strandings and could be the reason seasonality was not significant in our sea turtle data.


Deceased Fin Whale - Joshua Meza-Fidalgo, 2020

Deceased Fin Whale - Joshua Meza-Fidalgo, 2020

References

Data: https://seamap.env.duke.edu/dataset/945

  1. Halpin, P.N., A.J. Read, E. Fujioka, B.D. Best, B. Donnelly, L.J. Hazen, C. Kot, K. Urian, E. LaBrecque, A. Dimatteo, J. Cleary, C. Good, L.B. Crowder, and K.D. Hyrenbach. 2009. OBIS-SEAMAP: The world data center for marine mammal, sea bird, and sea turtle distributions. Oceanography. 22(2):104-115.

  2. Smith, A. 2014. Mystic Aquarium’s marine mammal and sea turtle stranding data 1976-2011. Data downloaded from OBIS-SEAMAP (http://seamap.env.duke.edu/dataset/945) on 2022-04-02.

Photos: Special thanks to Joshua Meza-Fidalgo for providing whale photos for our report.